A security researcher discovered that a stolen canary token was used to make 85 calls to AWS services, originating from 84 different home internet lines across 28 countries. This sophisticated residential proxy setup made the malicious activity appear as legitimate household traffic, evading standard security monitoring. The ultimate goal of these calls was to access AI models. AI
IMPACT Highlights a new attack vector targeting AI models via compromised credentials and sophisticated proxy networks.
RANK_REASON The item describes a sophisticated method for accessing AI models using a stolen canary token and a residential proxy network, which is a novel attack vector.
Read on Mastodon — mastodon.social →
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →